All research systems
Agentic FabricActive

What does an orchestration layer owe the guarantees below it?

Motivation

Orchestration frameworks tend to treat memory, authorization, and telemetry as optional plugins. If routine execution can bypass them, the guarantees they provide are advisory. The fabric should make enforcement structural rather than opt-in.

Architecture

  • Routine definitions with explicit pre- and post-conditions
  • Enforcement hooks so grants and memory writes cannot be skipped
  • Model fallbacks that prefer refusal over fabricated success
  • Deterministic replay of a routine from its recorded trace

Evaluation methodology

Supporting infrastructure. Evaluation in progress: measuring hook-bypass rate and replay fidelity across routines.

Results

Evaluation in progress

Limitations

  • Replay fidelity is untested against non-deterministic external tools

Current status

Orchestration stack running the routines used by the other projects, with enforcement hooks wired to Agent Guard and MemKit.

Roadmap

  • Replay-based regression testing for routines
  • Cross-run comparison of decision paths

Future research

  • Composition semantics for enforced routines
  • Failure containment across multi-agent routines